Liger-Kernel
Efficient Triton Kernels for LLM Training
About
Liger Kernel is a collection of Triton kernels designed specifically for LLM training. It can effectively increase multi-GPU training throughput by 20% and reduces memory usage by 60%. We have implemented Hugging Face Compatible RMSNorm, RoPE, SwiGLU, CrossEntropy, FusedLinearCrossEntropy, and more to come. The kernel works out of the box with Flash Attention, PyTorch FSDP, and Microsoft DeepSpeed. We welcome contributions from the community to gather the best kernels for LLM training.
Open Source Health
- Stars
- 6,610
- Forks
- 599
- License
- BSD-2-Clause
- Last commit
- 28 days ago
Related Categories
Vendor
Publisher of LinkedIn Sales Navigator and LinkedIn Sales Navigator
Quick Links
Open Source
More by Linkedin
Related Products
Peft
🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.
Top category match
Sophia Script For Windows
zap: The most powerful PowerShell module for fine-tuning Windows 10 & Windows 11 on GitHub
Top category match
KTransformers
A Flexible Framework for Experiencing Heterogeneous LLM Inference/Fine-tune Optimizations
Top category match
LongWriter
[ICLR 2025] LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs
Top category match
Regress Lm
Library for sequence-to-sequence numeric prediction, applicable to any tokenizable input, and allows pretraining and fine-tuning over multiple tasks.
Top category match